Research On The Digital Watermark Technology To Improve The Accuracy Of The Semi-fragile For Tamper Detection

As the development of digital media technology and the popularity of computer network, the communication and transmission of digital media information become more and more convenient. But this convenience also brings many problems, such as unauthorized copied, malicious tampering, and even forging of the digital media information. Therefore, the importance of the copyright protection of digital media is becoming increasingly prominent. On this background, the digital watermark technology develops. Presently, the digital watermark is regarded as the synonym of digital media protection. This dissertation introduces and explains the features and the sort of the digital watermark, and focus on studying the characteristic, theoretical framework and algorithm of the semi-fragile watermark in detail. The main contents of this dissertation are as follows:(1) First this thesis analyzes and verifies the semi-fragile characteristics of Hu moments. By observing the images’ Hu moments before and after being tampered with maliciously and non-maliciously, it’s found that the Hu moments show vulnerability to the malicious tampering, while robustness to the non-malicious tampering. On this basis, variety and mean square difference of the Hu moments before and after the images are tampered with can be used to put forward a method to authenticate the tampering. Experiments show: the accuracy of the tamper authentication can reach 90% or above. Secondly, when studying the coefficients of the WBCT transform domain: the relationship between the sub-band coefficients in different directions of the secondary high frequency shows semi-fragile. By using this relationship, an algorithm of semi-fragile watermark is designed in order to embed and extract the watermarks by modulating the coefficients. Finally, the combination of the authentication method based on Hu moments and the fragile watermark’s algorithm based on WBCT transform proposed above can realized the purpose of authentication and detection of images.(2) this thesis analyzes and verifies that there are defects in the semi-fragile watermark algorithm of the fixed threshold setting when detecting the tampering. To solve this problem, a method is presented to improve the precision of tamper detection, which can dynamically adjust the threshold according to the compression ratio in the condition of JPEG2000 compression. Secondly, based on the distribution of the JPEG2000 coefficients, by using the generalized Benford’s rule, an effective estimate algorithm of the JPEG2000 compression ratio is put forward, and it’s verified by experiments that the estimation precision of this algorithm can reach as high as 95% or above. The simulation result shows that, by adjusting the relationship between the JPEG2000 compression ratio and the threshold, the precision of the tamper detection can be greatly improved.(3) In order to combine the innovative theory presented in this dissertation and the practice application, mixed programming of MATLAB and C++ can be used to design and implement the watermark system, which will further verify the effectiveness of the algorithm.